In this section I import the data and prepare it for analysis.
Code
```{r setup}#| warning: false#| message: falselibrary(tidyverse)library(plotly)library(dplyr)library(tidyr)library(knitr)library(table1) #Create HTML Tables of Descriptive Statistics https://cran.r-project.org/web/packages/table1/vignettes/table1-examples.html#library(OMTM1) #https://github.com/schildjs/OMTM1/library(Hmisc)library(rms) # Regression Modeling Strategies by Frank https://cran.r-project.org/web/packages/rms/index.htmllibrary(modelsummary) #Summary Tables and Plots for Statistical Models and Data: Beautiful, Customizable, and Publication-Ready https://cran.r-project.org/web/packages/modelsummary/index.htmllibrary(scales) # The scales packages provides the internal scaling infrastructure used by ggplot2, and gives you tools to override the default breaks, labels, transformations and palettes. https://scales.r-lib.orglibrary(viridis) #colorslibrary(cowplot) #allows me to use plotgridlibrary(gridExtra) #adding tables to plotslibrary(visdat) #shows missing datalibrary(GGally) #makes pairs plotslibrary(sandwich) #for robust standard errorslibrary(gtsummary) #makes the tables look nicersetwd("/Users/lisalevoir/BIOS7351_Collab/data_project2") #this line I would need to run in the consoleknitr::opts_knit$set(root.dir ="/Users/lisalevoir/BIOS7351_Collab/github/BIOS_Collaboration/project_2_analysis") #now I set global options for knitting, I also had to toggle global options > R Markdown > evaluate chunks in current directory#import the datadat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/combined_data_203.csv")#to compare that the merging went as expectedVUMSdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_VUMS.csv")HMSdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_HMS.csv")UVAdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_UVA.csv")```
1.0.1 Inclusion/Exclusion
Criteria to exclude students who most likely took a scored exam:
Any PhD students (n = 2)
Any 5th year program students (n = 19)
M4 students at Vanderbilt (n = 5)
Students who did not complete either Step survey (n = 2)
Students who specifically stated they took a scored Step 1 (n=1)
Based on our criteria we would exclude record IDs:
[1] "uworld_percent_step1_1" "uworld_percent_step1_2"
[1] "amboss_percent_step1_1" "amboss_percent_step1_2"
[1] "length_step1_1" "length_step1_2"
[1] "practicetest_step1_1" "practicetest_step1_2"
[1] "full_test_practice_step1_1" "full_test_practice_step1_2"
[1] "push_step1_1" "push_step1_2"
[1] "push_practice_test_step1_1" "push_practice_test_step1_2"
[1] "push_nbme_practice_score_step1_1" "push_nbme_practice_score_step1_2"
[1] "push_uw_practice_score_step1_1" "push_uw_practice_score_step1_2"
[1] "final_nbme_practice_score_step1_1" "final_nbme_practice_score_step1_2"
[1] "final_uw_practice_score_step1_1" "final_uw_practice_score_step1_2"
[1] "score_step1_1" "score_step1_2"
[1] "other_resources_step1_1" "other_resources_step1_2"
[1] "other_step1_1" "other_step1_2"
[1] "changes_step1_1" "changes_step1_2"
[1] "Above is a list of columns I have combined for you. Hope it looks right!"
[1] "uworld_percent_step2_1" "uworld_percent_step2_2"
[1] "amboss_percent_step2_1" "amboss_percent_step2_2"
[1] "length_step2_1" "length_step2_2"
[1] "practicetest_step2_1" "practicetest_step2_2"
[1] "full_test_practice_step2_1" "full_test_practice_step2_2"
[1] "practice_score_step2_1" "practice_score_step2_2"
[1] "practice_test_step2_1" "practice_test_step2_2"
[1] "score_step2_1" "score_step2_2"
[1] "target_score_step2_1" "target_score_step2_2"
[1] "Above is a list of columns I have combined for you. Hope it looks right!"
There were 203 participants after the the clients did data exclusion by hand, as compared to the total survey size which was 182 unique individuals included in the step 1 data.
These are the record IDs that were excluded by the collaborator: VUSM 23, VUSM 60, HMS 37, HMS 41, HMS 49, UVASOM 33, UVASOM 80, UVASOM 84.
For our analysis, I originally included everyone from the 3 school specific data sets. Then, we flagged individuals to remove and I remade the source data set. These are the records IDs that we decided to exclude:
VU record ids: 23, 26, 39, 40, 54, 3, 8, 12, 49, 60, 62, 64
HMS record ids: 1, 21, 28, 30, 34, 37, 39, 41, 44, 47, 49, 61
UVA record ids: 33, 47, 80, 81, 83
Below are tables for Step 1 and Step 2 response inclusion by school:
Code
kable(table(step1_complete$schoolid), caption ="Number of Step 1 responses included by school")
Number of Step 1 responses included by school
Var1
Freq
HMS
50
UVa
79
VU
53
Code
kable(table(step2_complete$schoolid), caption ="Number of Step 2 responses included by school")
Number of Step 2 responses included by school
Var1
Freq
HMS
39
UVa
65
VU
34
Notice that unfortunately for Step 2 analysis, we had to drop 44 individuals who did not report a step 2 score. These raw counts and frequencies of people who did not give a step 2 score (and are therefor not eligible for analysis) are listed by institution below.
Code
# describing the missingnesskable(table(drop_these_with_no_outcome$schoolid), caption ="Number of missing Step 2 scores by institution")
Number of missing Step 2 scores by institution
Var1
Freq
HMS
11
UVa
14
VU
19
Code
num <-as.vector(table(drop_these_with_no_outcome$schoolid))denom <-as.vector(table(step2_complete$schoolid))nonresponse_freq <-setNames(c(round(num/denom, 3)), c(names(table(step2_complete$schoolid))))kable(nonresponse_freq, caption ="frequency of missing step 2 scores by instition")
frequency of missing step 2 scores by instition
x
HMS
0.282
UVa
0.215
VU
0.559
We also find it helpful to visually profile the missingness in this graphic below. Some of this missingness is due to questions being hierarchical.
Note, we plan to use the robust/sandwich variance estimator for regression models. One inclusion criteria is that the outcome variable (“Y”) must be available for a subject to be included in the analysis question (ie. if they did not report a step 2 score, we won’t perform relevant step 2 analysis on them).
summary(exam_order_mod) #model summary for reporting
Call:
lm(formula = score_step2 ~ which_exam_first + school_factor,
data = step2_complete)
Residuals:
Min 1Q Median 3Q Max
-112.100 -5.178 2.767 8.761 21.900
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 261.3383 5.2033 50.225 <2e-16 ***
which_exam_first 0.8952 4.5256 0.198 0.844
school_factorUVA -4.0290 5.0342 -0.800 0.425
school_factorVUMS -0.9797 3.5128 -0.279 0.781
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 14.69 on 134 degrees of freedom
Multiple R-squared: 0.01018, Adjusted R-squared: -0.01198
F-statistic: 0.4593 on 3 and 134 DF, p-value: 0.7112
Based on the model output (p values), there doesn’t appear to be any signifigant associations between exam order and the score for Step 2. Since the R^2 value is essentially 0, I conclude that there was no effect of exam order on step 2 scores.
Again, we cannot analyze Step 1 scores since all respondents reported passing.
Based on our SAP, if there are any covariates with more than 30% of responses missing, we will drop that variable or populate it with 0, depending on context. For example, the percent of Amboss questions completed will be filled with 0 for people who didn’t answer that they used Amboss since it seems safe to assume they didn’t complete any of the Amboss questions. For people who did select that they used Amboss but didn’t answer a percentage will be treated as missing
If less than 30% are missing, I may consider performing bootstrap sampling of known values to replace missing values.
After accounting for missingness, I will assess for co-linearity of the predictors (ie. correlation) using VIF. If there is high co-linearity, we will use LASSO to perform variable selection. If there is no evidence of concerning levels of colinearity, I will proceed with linear regression.
Code
#here I am creating the cleaned Uworld and Amboss variables. if they did not check that they used it, the percentage is set to 0.# if they did use it then the next ifelse checks if there is an NA where there should be a percent# if there is an NA where there should be a percent, we maintain this NA. If there is not, then we can report the percentage.# same process for uworldstep2_complete <- step2_complete %>%mutate(amboss_clean =ifelse(I.amboss ==0, 0, ifelse(is.na(step2_complete$amboss_percent_step2) ==TRUE, NA, step2_complete$amboss_percent_step2)),uworld_clean =ifelse(I.uworld ==0, 0, ifelse(is.na(step2_complete$uworld_percent_step2 ) ==TRUE, NA, step2_complete$uworld_percent_step2)))######## profile missingness in the step 2 data and addresspercents_missing <-round(colSums(is.na(step2_complete))/nrow(step2_complete), 3)*100kable(percents_missing, caption ="Percent missing observations for pooled Step 2 survey")
Percent missing observations for pooled Step 2 survey
x
record_id
0.0
schoolid
0.0
exam_order
0.0
uworld_percent_step2
1.4
amboss_percent_step2
76.8
length_step2
1.4
practicetest_step2
0.0
full_test_practice_step2
0.0
practice_score_step2
29.7
practice_test_step2
29.0
score_step2
0.0
target_score_step2
0.0
I.uworld
0.0
I.firstaid
0.0
I.anki
0.0
I.sketchy
0.0
I.amboss
0.0
I.pathoma
0.0
I.boardsbeyond
0.0
I.other
0.0
order_factor
0.0
which_exam_first
0.0
school_factor
0.0
amboss_clean
9.4
uworld_clean
1.4
Code
#inspecting percent missing, it seems like most responses are now complete enough for analysisstep2_complete$amboss_percent_step2 <-ifelse(is.na(step2_complete$amboss_percent_step2) ==TRUE, 0, step2_complete$amboss_percent_step2)class(step2_complete$practicetest_step2) <-"integer"step2_complete[,"on_target"] <-factor(step2_complete$target_score_step2, levels =c(1,2,3), labels =c("at target", "above target", "below target"))step2_complete[, "practice_test_2_clean"] <-ifelse(is.na(step2_complete$practicetest_step2) ==TRUE, NA, substr(step2_complete$practicetest_step2, start =1, stop =1))step2_complete[, "number_of_practice_tests"] <-as.numeric(step2_complete$practice_test_2_clean)
Multiple linear regression with:
Y = Step 2 score
X1 = % UWorld
X2 = % Amboss
X3 = length study
X4 = # of practice tests
X5 = full test day (yes/no code as binary)
X6 = final practice score (however, is there some conversion between U World and Amboss - Jeffrey is looking into this)
Z = School (need to adjust for this)
Careful to note that not all cases are complete - for example there are 399 responses in the complete step 2 dataset, of which for the number of practice tests taken, 108 are missing and 291 have a response recorded.
Below I report the model results, sandqich variance, and VIF for step 2 scores model.
mod_step2_scores2 <-lm(score_step2 ~ uworld_percent_step2 + amboss_percent_step2 + length_step2 + simulate_full_practice + practice_score_step2 + practice_test_2_clean + school_factor, data = step2_complete)summary(mod_step2_scores2) #we need to talk about these model results
I looked into the difference between practice_test_step2 (the final practice test I took before my exam was… 8 options) and practicetest_step2 (text response of how many practice tests did you take before step 2). I decided not to include either in the model.
Note
We need to talk about the model results between the two ways I coded the variables for # of practice tests
Here, I will perform logistic regression with
Y = yes or no (1 = yes, 2 = no for “push_step1”)
There may not be sufficient data on this since only 20 people responded that they decided to push back Step 1. The factors that were measured are:
push remember step1 (1 = I only remember the form name, 2 = I only remember the score, 3 = I remember the form name and the score, 4 = I don’t remember either) - we decided not to include this variable (can change later if desired)
push score only step 1 (1 = NBME, 2 = Uworld)
push practice test step 1 ( 1 - 8 listing various exams)
push nbme practice score (from 0 to 100%)
push uw practice score (from 180 to 300)
Listing variables by name and if I have included them:
“push_step1_1” yes
“push_remember_step1_1” not included b/c a precursor question
“push_score_only_step1_1” not currently included but could be
“push_practice_test_step1_1” yes
“push_nbme_practice_score_step1_1” yes
“push_uw_practice_score_step1_1” yes
Code
step1_complete$push_step1 <-ifelse(step1_complete$push_step1 ==2|is.na(step1_complete$push_step1) ==TRUE, 2, 1) #recording the NA's to be "No" (they did not push back step 1)step1_complete$push_step1_label <-factor(step1_complete$push_step1, levels =c(1, 2), labels =c("Yes", "No")) #making a nice descriptive labeldid_push_df <- step1_complete %>%filter(push_step1_label =="Yes")dat %>%filter(!is.na(push_remember_step1_1))
X record_id school timestamp year exam_order step_spacing
1 9 7 VUSM 2/8/23 9:31 3 1 2
2 12 12 VUSM 2/8/23 10:05 5 1 1
3 23 39 VUSM 2/16/23 20:12 6 1 3
4 29 51 VUSM 2/20/23 12:57 3 1 3
5 47 9 VUSM 2/8/23 9:53 3 3 3
6 50 20 VUSM 2/8/23 11:05 3 1 3
7 60 57 VUSM 2/21/23 10:19 3 3 3
8 75 12 UVASOM 2023-08-17 11:31:21 7 1 2
9 129 67 UVASOM 2023-09-07 09:48:22 7 1 2
10 141 79 UVASOM 2023-09-07 16:02:17 7 1 2
11 174 30 HMS 2023-08-23 18:50:56 4 3 2
12 196 55 HMS 2023-10-04 08:31:07 7 1 2
13 199 58 HMS 2023-10-04 15:27:27 7 1 2
step_spacing_months survey_complete step_1_first_timestamp
1 NA 2 2/8/23 9:33
2 NA 2 2/8/23 10:11
3 5 2 2/16/23 20:14
4 3 2 2/20/23 12:59
5 7 2 2/8/23 9:56
6 5 2 2/8/23 11:19
7 6 2 2/21/23 10:26
8 NA 2 2023-08-17 11:32:55
9 NA 2 2023-09-07 09:49:48
10 NA 2 2023-09-07 16:04:58
11 NA 2 2023-08-23 19:10:48
12 NA 2 2023-10-04 08:37:24
13 NA 2 2023-10-04 15:29:01
resources_step1_1___1 resources_step1_1___2 resources_step1_1___3
1 1 1 0
2 1 1 1
3 1 1 1
4 1 1 1
5 1 1 0
6 1 1 1
7 0 0 1
8 1 1 1
9 1 1 0
10 1 1 0
11 1 1 0
12 1 1 1
13 1 1 0
resources_step1_1___4 resources_step1_1___5 resources_step1_1___6
1 0 0 0
2 1 0 0
3 1 0 1
4 1 0 1
5 1 0 0
6 1 0 0
7 1 0 0
8 0 1 0
9 0 0 0
10 0 0 0
11 1 0 0
12 1 1 1
13 0 0 0
resources_step1_1___7 resources_step1_1___8 other_resources_step1_1
1 0 0 <NA>
2 0 0 <NA>
3 1 0 <NA>
4 0 0 <NA>
5 0 0 <NA>
6 0 0 <NA>
7 0 0 <NA>
8 0 0 <NA>
9 0 0 <NA>
10 0 0 <NA>
11 0 1 Osmosis
12 1 0
13 0 0
other_step1_1 uworld_percent_step1_1 uworld_step1_1 first_aid_step1_1
1 <NA> NA 3 1
2 <NA> 60 3 1
3 <NA> 80 1 1
4 <NA> 65 1 1
5 <NA> 50 1 1
6 <NA> 69 1 1
7 <NA> NA NA NA
8 <NA> 75 1 1
9 <NA> 70 3 2
10 <NA> 35 3 1
11 Yes 30 3 2
12 90 3 1
13 86 1 1
anki_step1_1 anki_use_step1_1 anki_details_step1_1___1
1 NA NA 0
2 1 1 0
3 1 3 0
4 1 3 1
5 NA NA 0
6 1 1 1
7 1 3 1
8 1 1 0
9 NA NA 0
10 NA NA 0
11 NA NA 0
12 1 3 0
13 NA NA 0
anki_details_step1_1___2 anki_details_step1_1___3 anki_details_step1_1___4
1 0 0 0
2 0 1 0
3 1 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 1 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 1 0
13 0 0 0
anki_details_step1_1___5 anki_details_step1_1___6 anki_details_step1_1___7
1 0 0 0
2 0 1 0
3 0 1 0
4 0 1 0
5 0 0 0
6 0 0 1
7 1 1 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 1 0
13 0 0 0
sketchy_step1_1 sketchy_details_step1_1___1 sketchy_details_step1_1___2
1 NA 0 0
2 1 1 1
3 1 1 1
4 1 1 1
5 1 1 1
6 1 0 1
7 1 1 1
8 NA 0 0
9 NA 0 0
10 NA 0 0
11 1 0 0
12 1 1 1
13 NA 0 0
sketchy_details_step1_1___3 sketchy_details_step1_1___4
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step1_1___5 sketchy_details_step1_1___6
1 0 0
2 0 1
3 0 0
4 0 0
5 0 0
6 0 0
7 0 1
8 0 0
9 0 0
10 0 0
11 1 0
12 0 1
13 0 0
sketchy_details_step1_1___7 amboss_details_step1_1___1
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
amboss_details_step1_1___2 amboss_details_step1_1___3 amboss_library_step1_1
1 0 0 NA
2 0 0 NA
3 0 0 NA
4 0 0 NA
5 0 0 NA
6 0 0 NA
7 0 0 NA
8 1 0 1
9 0 0 NA
10 0 0 NA
11 0 0 NA
12 0 1 1
13 0 0 NA
amboss_percent_step1_1 amboss_amount_step1_1 pathoma_step1_1 bnb_step1_1
1 NA NA NA NA
2 NA NA NA NA
3 NA NA 1 1
4 NA NA 1 NA
5 NA NA NA NA
6 NA NA NA NA
7 NA NA NA NA
8 NA NA NA NA
9 NA NA NA NA
10 NA NA NA NA
11 NA NA NA NA
12 32 3 1 1
13 NA NA NA NA
bnb_how_step1_1___1 bnb_how_step1_1___2 bnb_how_step1_1___3
1 0 0 0
2 0 0 0
3 1 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 1 0 0
13 0 0 0
bnb_how_step1_1___4 length_step1_1 practicetest_step1_1
1 0 4 4
2 0 3 3
3 0 6 6
4 0 4 4
5 0 3 5
6 0 6 6
7 0 6 5
8 0 4 4
9 0 4 3
10 0 4 3
11 0 3 4
12 0 5 3
13 0 5 5
practice_test_amount_step1_1 full_test_practice_step1_1 push_step1_1
1 3 2 1
2 3 1 1
3 2 4 1
4 1 3 1
5 3 1 1
6 3 3 1
7 3 4 1
8 3 3 1
9 1 3 1
10 3 1 1
11 3 4 1
12 3 3 1
13 1 1 1
push_remember_step1_1 push_score_only_step1_1 push_practice_test_step1_1
1 3 NA 4
2 4 NA NA
3 4 NA NA
4 2 1 NA
5 4 NA NA
6 4 NA NA
7 3 NA 3
8 4 NA NA
9 2 1 NA
10 4 NA NA
11 2 1 NA
12 2 1 NA
13 4 NA NA
push_nbme_practice_score_step1_1 push_uw_practice_score_step1_1
1 65 NA
2 NA NA
3 NA NA
4 70 NA
5 NA NA
6 NA NA
7 80 NA
8 NA NA
9 87 NA
10 NA NA
11 83 NA
12 30 NA
13 NA NA
final_practice_step1_1 final_practice_test_step1_1
1 1 6
2 0 NA
3 0 NA
4 1 7
5 0 NA
6 0 NA
7 1 2
8 1 7
9 1 7
10 1 6
11 1 5
12 0 NA
13 0 NA
final_nbme_practice_score_step1_1 final_uw_practice_score_step1_1
1 68 NA
2 NA NA
3 NA NA
4 95 NA
5 NA NA
6 NA NA
7 97 NA
8 97 NA
9 98 NA
10 99 NA
11 92 NA
12 NA NA
13 NA NA
score_step1_1 retake_step1_1 retake_pass_step1_1 study_amount_step1_1
1 1 NA NA 3
2 1 NA NA 1
3 1 NA NA 1
4 1 NA NA 1
5 1 NA NA 2
6 1 NA NA 2
7 1 NA NA 2
8 1 NA NA 1
9 1 NA NA 1
10 1 NA NA 3
11 1 NA NA 1
12 1 NA NA 1
13 1 NA NA 2
changes_step1_1
1
2 Didn't know how to note this above, but I did step 1 anki for the entirety of my last clerkship then spent 2 weeks of dedicated studying
3
4
5 I would have started reviewing material (i.e. watching B&B, Sketchy pharm) before dedicated started so that I could devote my dedicated time almost exclusively to UWorld questions
6 I ended up taking 12 weeks. M1 was terrible for me and I didn't know how to study. So, I would go back and study more for it during M1 or M2 but even 12 weeks was pushing it for me to make up the deficit I was in
7 I wish I would have started the premade anki deck during M1 instead of at the start of M2. I finished about 50% of STEP 1 cards and 75% of STEP 2 cards by the time I took Step 1 and I think it would have been much quicker to a passing score if i had a higher percentage of STEP 1 cards mastered. I did not use UWorld at all because I found that reviewing missed questions didn't help me at all. I never remembered the info or explanations unless I memorized it with anki first. Once I had things memorized, I had no issues applying the knowledge in exam form. The only questions I missed were ones that I had never seen on anki.
8 I wish it was after pre-clinical year and not after clinical year.
9 I wish we had completed step 1 post pre clerkship with dedicated time.
10 I would focus on pathophysiology section of uworld
11 I wish I didn't listen to the advice/guidance that my school (HMS) gave me. They really emphasized use of First Aid and at first I followed their guidance (because they shared so little guidance/support overall) even though I have never been a textbook reader, and so I wasted a lot of time studying via methods that were not compatible with my learning style. I also wish I had the opportunity to study with peers or through an HMS class because I also learn much better by talking through concepts with others, but there were no classes/study groups/tutoring or really any support offered by HMS. Given how much we pay for tuition to HMS, I think we should be able to opt-in or sign up for a Step 1 prep course - at a minimum this would provide some structure to an otherwise isolating and irritating experience.
12
13 Spent more time studying during pre-clinical years
step_1_first_complete step_2_first_timestamp resources_step2_1___1
1 2 0
2 2 0
3 2 0
4 2 0
5 2 0
6 2 0
7 2 0
8 2 0
9 2 0
10 2 0
11 2 0
12 2 0
13 2 0
resources_step2_1___2 resources_step2_1___3 resources_step2_1___4
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
resources_step2_1___5 resources_step2_1___6 resources_step2_1___7
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
resources_step2_1___8 other_resources_step2_1 other_step2_1
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 0
9 0
10 0
11 0 <NA> <NA>
12 0 <NA> <NA>
13 0 <NA> <NA>
uworld_percent_step2_1 uworld_step2_1 first_aid_step2_1 anki_step2_1
1 NA NA NA NA
2 NA NA NA NA
3 NA NA NA NA
4 NA NA NA NA
5 NA NA NA NA
6 NA NA NA NA
7 NA NA NA NA
8 NA NA NA NA
9 NA NA NA NA
10 NA NA NA NA
11 NA NA NA NA
12 NA NA NA NA
13 NA NA NA NA
anki_use_step2_1 anki_details_step2_1___1 anki_details_step2_1___2
1 NA 0 0
2 NA 0 0
3 NA 0 0
4 NA 0 0
5 NA 0 0
6 NA 0 0
7 NA 0 0
8 NA 0 0
9 NA 0 0
10 NA 0 0
11 NA 0 0
12 NA 0 0
13 NA 0 0
anki_details_step2_1___3 anki_details_step2_1___4 anki_details_step2_1___5
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
anki_details_step2_1___6 anki_details_step2_1___7 sketchy_step2_1
1 0 0 NA
2 0 0 NA
3 0 0 NA
4 0 0 NA
5 0 0 NA
6 0 0 NA
7 0 0 NA
8 0 0 NA
9 0 0 NA
10 0 0 NA
11 0 0 NA
12 0 0 NA
13 0 0 NA
sketchy_details_step2_1___1 sketchy_details_step2_1___2
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step2_1___3 sketchy_details_step2_1___4
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step2_1___5 sketchy_details_step2_1___6
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step2_1___7 amboss_details_step2_1___1
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
amboss_details_step2_1___2 amboss_details_step2_1___3 amboss_library_step2_1
1 0 0 NA
2 0 0 NA
3 0 0 NA
4 0 0 NA
5 0 0 NA
6 0 0 NA
7 0 0 NA
8 0 0 NA
9 0 0 NA
10 0 0 NA
11 0 0 NA
12 0 0 NA
13 0 0 NA
amboss_percent_step2_1 amboss_amount_step2_1 pathoma_step2_1 bnb_step2_1
1 NA NA NA NA
2 NA NA NA NA
3 NA NA NA NA
4 NA NA NA NA
5 NA NA NA NA
6 NA NA NA NA
7 NA NA NA NA
8 NA NA NA NA
9 NA NA NA NA
10 NA NA NA NA
11 NA NA NA NA
12 NA NA NA NA
13 NA NA NA NA
bnb_how_step2_1___1 bnb_how_step2_1___2 bnb_how_step2_1___3
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
bnb_how_step2_1___4 length_step2_1 practicetest_step2_1
1 0 NA <NA>
2 0 NA <NA>
3 0 NA <NA>
4 0 NA <NA>
5 0 NA <NA>
6 0 NA <NA>
7 0 NA <NA>
8 0 NA
9 0 NA
10 0 NA
11 0 NA <NA>
12 0 NA <NA>
13 0 NA <NA>
practice_test_amount_step2_1 full_test_practice_step2_1
1 NA NA
2 NA NA
3 NA NA
4 NA NA
5 NA NA
6 NA NA
7 NA NA
8 NA NA
9 NA NA
10 NA NA
11 NA NA
12 NA NA
13 NA NA
final_practice_step2_1 practice_test_step2_1 practice_score_step2_1
1 NA NA NA
2 NA NA NA
3 NA NA NA
4 NA NA NA
5 NA NA NA
6 NA NA NA
7 NA NA NA
8 NA NA NA
9 NA NA NA
10 NA NA NA
11 NA NA NA
12 NA NA NA
13 NA NA NA
score_step2_1 retake_step2_1 retake_pass_step2_1 target_score_step2_1
1 NA NA NA
2 NA NA NA
3 NA NA NA
4 NA NA NA
5 NA NA NA
6 NA NA NA
7 NA NA NA
8 <NA> NA NA NA
9 <NA> NA NA NA
10 <NA> NA NA NA
11 <NA> NA NA NA
12 <NA> NA NA NA
13 <NA> NA NA NA
study_amount_step2_1 changes_step2_1 step_2_first_complete
1 NA 0
2 NA 0
3 NA 0
4 NA 0
5 NA 0
6 NA 0
7 NA 0
8 NA 0
9 NA 0
10 NA 0
11 NA 0
12 NA 0
13 NA 0
step_1_second_timestamp resources_step1_2___1 resources_step1_2___2
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
resources_step1_2___3 resources_step1_2___4 resources_step1_2___5
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
resources_step1_2___6 resources_step1_2___7 resources_step1_2___8
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
other_resources_step1_2 other_step1_2 uworld_percent_step1_2
1 <NA> <NA> NA
2 <NA> <NA> NA
3 <NA> <NA> NA
4 <NA> <NA> NA
5 <NA> <NA> NA
6 <NA> <NA> NA
7 <NA> <NA> NA
8 NA
9 NA
10 NA
11 <NA> <NA> NA
12 <NA> <NA> NA
13 <NA> <NA> NA
first_aid_step1_2 uworld_step1_2 anki_step1_2 anki_use_step1_2
1 NA NA NA NA
2 NA NA NA NA
3 NA NA NA NA
4 NA NA NA NA
5 NA NA NA NA
6 NA NA NA NA
7 NA NA NA NA
8 NA NA NA NA
9 NA NA NA NA
10 NA NA NA NA
11 NA NA NA NA
12 NA NA NA NA
13 NA NA NA NA
anki_details_step1_2___1 anki_details_step1_2___2 anki_details_step1_2___3
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
anki_details_step1_2___4 anki_details_step1_2___5 anki_details_step1_2___6
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 0 0
13 0 0 0
anki_details_step1_2___7 anki_details_step1_2___8 sketchy_step1_2
1 0 0 NA
2 0 0 NA
3 0 0 NA
4 0 0 NA
5 0 0 NA
6 0 0 NA
7 0 0 NA
8 0 0 NA
9 0 0 NA
10 0 0 NA
11 0 0 NA
12 0 0 NA
13 0 0 NA
sketchy_details_step1_2___1 sketchy_details_step1_2___2
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step1_2___3 sketchy_details_step1_2___4
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step1_2___5 sketchy_details_step1_2___6
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step1_2___7 sketchy_details_step1_2___8
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
amboss_details_step1_2___1 amboss_details_step1_2___2
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
amboss_details_step1_2___3 amboss_library_step1_2 amboss_percent_step1_2
1 0 NA NA
2 0 NA NA
3 0 NA NA
4 0 NA NA
5 0 NA NA
6 0 NA NA
7 0 NA NA
8 0 NA NA
9 0 NA NA
10 0 NA NA
11 0 NA NA
12 0 NA NA
13 0 NA NA
amboss_amount_step1_2 pathoma_step1_2 bnb_step1_2 bnb_how_step1_2___1
1 NA NA NA 0
2 NA NA NA 0
3 NA NA NA 0
4 NA NA NA 0
5 NA NA NA 0
6 NA NA NA 0
7 NA NA NA 0
8 NA NA NA 0
9 NA NA NA 0
10 NA NA NA 0
11 NA NA NA 0
12 NA NA NA 0
13 NA NA NA 0
bnb_how_step1_2___2 bnb_how_step1_2___3 bnb_how_step1_2___4 length_step1_2
1 0 0 0 NA
2 0 0 0 NA
3 0 0 0 NA
4 0 0 0 NA
5 0 0 0 NA
6 0 0 0 NA
7 0 0 0 NA
8 0 0 0 NA
9 0 0 0 NA
10 0 0 0 NA
11 0 0 0 NA
12 0 0 0 NA
13 0 0 0 NA
practicetest_step1_2 practice_test_amount_step1_2 full_test_practice_step1_2
1 <NA> NA NA
2 <NA> NA NA
3 <NA> NA NA
4 <NA> NA NA
5 <NA> NA NA
6 <NA> NA NA
7 <NA> NA NA
8 NA NA
9 NA NA
10 NA NA
11 <NA> NA NA
12 <NA> NA NA
13 <NA> NA NA
push_step1_2 push_remember_step1_2 push_score_only_step1_2
1 NA NA NA
2 NA NA NA
3 NA NA NA
4 NA NA NA
5 NA NA NA
6 NA NA NA
7 NA NA NA
8 NA NA NA
9 NA NA NA
10 NA NA NA
11 NA NA NA
12 NA NA NA
13 NA NA NA
push_practice_test_step1_2 push_nbme_practice_score_step1_2
1 NA NA
2 NA NA
3 NA NA
4 NA NA
5 NA NA
6 NA NA
7 NA NA
8 NA NA
9 NA NA
10 NA NA
11 NA NA
12 NA NA
13 NA NA
push_uw_practice_score_step1_2 final_practice_step1_2
1 NA NA
2 NA NA
3 NA NA
4 NA NA
5 NA NA
6 NA NA
7 NA NA
8 NA NA
9 NA NA
10 NA NA
11 NA NA
12 NA NA
13 NA NA
final_practice_test_step1_2 final_nbme_practice_score_step1_2
1 NA NA
2 NA NA
3 NA NA
4 NA NA
5 NA NA
6 NA NA
7 NA NA
8 NA NA
9 NA NA
10 NA NA
11 NA NA
12 NA NA
13 NA NA
final_uw_practice_score_step1_2 score_step1_2 retake_step1_2
1 NA NA NA
2 NA NA NA
3 NA NA NA
4 NA NA NA
5 NA NA NA
6 NA NA NA
7 NA NA NA
8 NA NA NA
9 NA NA NA
10 NA NA NA
11 NA NA NA
12 NA NA NA
13 NA NA NA
retake_pass_step1_2 study_amount_step1_2 changes_step1_2
1 NA NA
2 NA NA
3 NA NA
4 NA NA
5 NA NA
6 NA NA
7 NA NA
8 NA NA
9 NA NA
10 NA NA
11 NA NA <NA>
12 NA NA <NA>
13 NA NA <NA>
step_1_second_complete step_2_second_timestamp resources_step2_2___1
1 0 2/8/23 9:35 1
2 0 2/8/23 10:12 1
3 0 2/16/23 20:15 1
4 0 2/20/23 13:01 1
5 0 0
6 0 0
7 0 0
8 0 0
9 0 2023-09-07 09:50:43 1
10 0 2023-09-07 16:06:06 1
11 0 0
12 0 2023-10-04 08:42:48 1
13 0 2023-10-04 15:31:29 1
resources_step2_2___2 resources_step2_2___3 resources_step2_2___4
1 1 0 0
2 1 0 1
3 0 1 0
4 0 1 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 1 1
13 0 0 0
resources_step2_2___5 resources_step2_2___6 resources_step2_2___7
1 0 0 0
2 0 0 0
3 0 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 1 1 1
13 0 0 0
resources_step2_2___8
1 0
2 0
3 0
4 1
5 0
6 0
7 0
8 0
9 0
10 0
11 0
12 0
13 1
other_resources_step2_2 other_step2_2
1
2
3
4 Divine Intervention Podcasts Yes
5
6
7
8
9
10
11
12
13 Summary PDFs/content review I found online, youtube videos Yes
uworld_percent_step2_2 uworld_step2_2 first_aid_step2_2 anki_step2_2
1 70 1 2 NA
2 66 3 1 NA
3 65 1 NA 1
4 80 1 NA 1
5 NA NA NA NA
6 NA NA NA NA
7 NA NA NA NA
8 NA NA NA NA
9 50 1 NA NA
10 61 1 NA NA
11 NA NA NA NA
12 94 3 NA 1
13 71 1 NA NA
anki_use_step2_2 anki_details_step2_2___1 anki_details_step2_2___2
1 NA 0 0
2 NA 0 0
3 3 0 0
4 3 1 0
5 NA 0 0
6 NA 0 0
7 NA 0 0
8 NA 0 0
9 NA 0 0
10 NA 0 0
11 NA 0 0
12 3 0 0
13 NA 0 0
anki_details_step2_2___3 anki_details_step2_2___4 anki_details_step2_2___5
1 0 0 0
2 0 0 0
3 1 0 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 1 0 0
13 0 0 0
anki_details_step2_2___6 anki_details_step2_2___7 anki_details_step2_2___8
1 0 0 0
2 0 0 0
3 0 1 0
4 0 0 0
5 0 0 0
6 0 0 0
7 0 0 0
8 0 0 0
9 0 0 0
10 0 0 0
11 0 0 0
12 0 1 0
13 0 0 0
sketchy_step2_2 sketchy_details_step2_2___1 sketchy_details_step2_2___2
1 NA 0 0
2 1 1 0
3 NA 0 0
4 NA 0 0
5 NA 0 0
6 NA 0 0
7 NA 0 0
8 NA 0 0
9 NA 0 0
10 NA 0 0
11 NA 0 0
12 1 0 1
13 NA 0 0
sketchy_details_step2_2___3 sketchy_details_step2_2___4
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 1 0
13 0 0
sketchy_details_step2_2___5 sketchy_details_step2_2___6
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 0 0
13 0 0
sketchy_details_step2_2___7 sketchy_details_step2_2___8
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 1 0
13 0 0
amboss_details_step2_2___1 amboss_details_step2_2___2
1 0 0
2 0 0
3 0 0
4 0 0
5 0 0
6 0 0
7 0 0
8 0 0
9 0 0
10 0 0
11 0 0
12 1 1
13 0 0
amboss_details_step2_2___3 amboss_library_step2_2 amboss_percent_step2_2
1 0 NA NA
2 0 NA NA
3 0 NA NA
4 0 NA NA
5 0 NA NA
6 0 NA NA
7 0 NA NA
8 0 NA NA
9 0 NA NA
10 0 NA NA
11 0 NA NA
12 1 1 41
13 0 NA NA
amboss_amount_step2_2 pathoma_step2_2 bnb_step2_2 bnb_how_step2_2___1
1 NA NA NA 0
2 NA NA NA 0
3 NA NA NA 0
4 NA NA NA 0
5 NA NA NA 0
6 NA NA NA 0
7 NA NA NA 0
8 NA NA NA 0
9 NA NA NA 0
10 NA NA NA 0
11 NA NA NA 0
12 3 1 1 1
13 NA NA NA 0
bnb_how_step2_2___2 bnb_how_step2_2___3 bnb_how_step2_2___4 length_step2_2
1 0 0 0 4
2 0 0 0 3
3 0 0 0 5
4 0 0 0 5
5 0 0 0 NA
6 0 0 0 NA
7 0 0 0 NA
8 0 0 0 NA
9 0 0 0 4
10 0 0 0 4
11 0 0 0 NA
12 0 0 0 4
13 0 0 0 4
practicetest_step2_2 practice_test_amount_step2_2 full_test_practice_step2_2
1 4 1 3
2 3 3 4
3 5 2 4
4 4 1 3
5 NA NA
6 NA NA
7 NA NA
8 <NA> NA NA
9 3 1 3
10 3 1 1
11 <NA> NA NA
12 3 3 3
13 5 1 1
final_practice_step2_2 practice_test_step2_2 practice_score_step2_2
1 0 NA
2 1 7 251
3 1 3 243
4 1 7 242
5 NA NA
6 NA NA
7 NA NA
8 NA NA <NA>
9 0 NA <NA>
10 0 NA <NA>
11 NA NA <NA>
12 0 NA <NA>
13 1 5 235
score_step2_2 retake_step2_2 retake_pass_step2_2 target_score_step2_2
1 239 NA NA 3
2 250 NA NA 1
3 no score yet NA NA 1
4 244 NA NA 1
5 NA NA NA
6 NA NA NA
7 NA NA NA
8 NA NA NA
9 n/a NA NA 1
10 253 NA NA 1
11 <NA> NA NA NA
12 266 NA NA 2
13 230 NA NA 3
study_amount_step2_2
1 2
2 1
3 2
4 1
5 NA
6 NA
7 NA
8 NA
9 1
10 2
11 NA
12 1
13 2
changes_step2_2
1
2
3
4
5
6
7
8
9 Dedicated time exclusive to step 2 post clerkship with step 1 completed pre-clerkship.
10
11
12 Simulate a full day exam before taking the test. I had done half length practice tests leading up to it but found that I got fatigued on the test day for the last three sections.
13 Studied more for Step 1. Would have done Anki in the months leading up to the exam.
step_2_second_complete uniqueID
1 2 VUSM 7
2 2 VUSM 12
3 2 VUSM 39
4 2 VUSM 51
5 0 VUSM 9
6 0 VUSM 20
7 0 VUSM 57
8 0 UVASOM 12
9 2 UVASOM 67
10 2 UVASOM 79
11 0 HMS 30
12 2 HMS 55
13 2 HMS 58
Code
step1_complete$push_practice_test_step1 <-factor(step1_complete$push_practice_test_step1, levels =c(1:8), labels =c("NBME 25", "NBME 26", "NBME 27", "NBME 28", "NBME 29", "NBME 30", "UWorld 1", "UWorld 2"))units(step1_complete$push_uw_practice_score_step1) <-"score units"units(step1_complete$push_nbme_practice_score_step1) <-"percent"label(step1_complete$push_practice_test_step1) <-"exam that triggered pushing back"label(step1_complete$push_nbme_practice_score_step1) <-"NBME P(passing) that triggered pushing back"label(step1_complete$push_uw_practice_score_step1) <-"3 digit UWorld score that triggered pushing back"caption <-"Description of indiviuals that pushed back Step 1"table1(~ push_practice_test_step1 + push_uw_practice_score_step1 + push_nbme_practice_score_step1 |push_step1_label, data=step1_complete, topclass="Rtable1-zebra")
Yes (N=20)
No (N=162)
Overall (N=182)
exam that triggered pushing back
NBME 25
0 (0%)
0 (0%)
0 (0%)
NBME 26
0 (0%)
0 (0%)
0 (0%)
NBME 27
1 (5.0%)
0 (0%)
1 (0.5%)
NBME 28
1 (5.0%)
0 (0%)
1 (0.5%)
NBME 29
0 (0%)
0 (0%)
0 (0%)
NBME 30
0 (0%)
0 (0%)
0 (0%)
UWorld 1
0 (0%)
0 (0%)
0 (0%)
UWorld 2
1 (5.0%)
0 (0%)
1 (0.5%)
Missing
17 (85.0%)
162 (100%)
179 (98.4%)
3 digit UWorld score that triggered pushing back (score units)
Mean (SD)
180 (NA)
NA (NA)
180 (NA)
Median [Min, Max]
180 [180, 180]
NA [NA, NA]
180 [180, 180]
Missing
19 (95.0%)
162 (100%)
181 (99.5%)
NBME P(passing) that triggered pushing back (percent)
Mean (SD)
65.6 (18.3)
NA (NA)
65.6 (18.3)
Median [Min, Max]
67.5 [30.0, 87.0]
NA [NA, NA]
67.5 [30.0, 87.0]
Missing
12 (60.0%)
162 (100%)
174 (95.6%)
Descriptive statistics will be reported by school and total.
histogram of Step 2 scores
what resources are most widely used barplot
how long did students study barplot
number of practice tests histogram (among the people who answered the question)
## do a plot for all schools (in total) as well as Vanderbilt in particular## use a unified color schemeggplot(aes(x = score_step2), data= step2_complete) +geom_histogram() +theme_minimal() +xlab("Self Reported Step 2 Score") +ylab("Frequency") +labs(title ="Frequency of reported Step 2 Scores")
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
Code
#more data cleaning first first step 2 for the resouces they selected that they usedresources_step2 <-colSums(took_step2general[20:35])uworld <- resources_step2[1] + resources_step2[9]first_aid <- resources_step2[2] + resources_step2[10]anki <- resources_step2[3] + resources_step2[11]sketchy <- resources_step2[4] +resources_step2[12]amboss <- resources_step2[5] + resources_step2[13]pathoma <- resources_step2[6] + resources_step2[14]boards_and_beyond <- resources_step2[7] + resources_step2[15]other <- resources_step2[8] + resources_step2[16]totals <-c(uworld, first_aid, anki, sketchy, amboss, pathoma, boards_and_beyond, other)#same idea for step 1resources_step1 <-colSums(took_step1general[26:41])uworld1 <- resources_step1[1] + resources_step1[9]first_aid1 <- resources_step1[2] + resources_step1[10]anki1 <- resources_step1[3] + resources_step1[11]sketchy1 <- resources_step1[4] +resources_step1[12]amboss1 <- resources_step1[5] + resources_step1[13]pathoma1 <- resources_step1[6] + resources_step1[14]boards_and_beyond1 <- resources_step1[7] + resources_step1[15]other1 <- resources_step1[8] + resources_step1[16]totals1 <-c(uworld1, first_aid1, anki1, sketchy1, amboss1, pathoma1, boards_and_beyond1, other1)rdf <-data.frame(amount =c(totals, totals1), name =rep(names(totals), 2), step_exam =c(rep("Step 2", 8), rep("Step 1", 8) ))# http://www.sthda.com/english/wiki/ggplot2-barplots-quick-start-guide-r-software-and-data-visualization used this as a guide#what resources are most widely used? I want to sort this barplot in order of frequency but for some reason this wasn't working for me with reorder()ggplot(data=rdf, aes(x=name, y=amount, fill = step_exam)) +geom_bar(stat="identity", position=position_dodge())+theme_minimal() +coord_flip() +scale_fill_brewer(palette="Blues")
Code
## how long did the students study barplot# for step 1step1_complete$length_step1 <-factor(step1_complete$length_step1 , levels =c(1:6), labels =c("less than 1 week", "1-2 weeks", "3-4 weeks", "5-6 weeks", "7-8 weeks", "more than 8 weeks"))ggplot(data.frame(step1_complete), aes(x=length_step1)) +geom_bar() +theme_minimal() +xlab("Time") +ylab("Frequency") +labs(title ="How long did you study for Step 1 during a protected study period?")
Code
# for step 2ggplot(data.frame(step2_complete), aes(x=length_step2)) +geom_bar() +theme_minimal() +xlab("Time") +ylab("Frequency") +labs(title ="How long did you study for Step 2 during a protected study period?")
Code
## number of practice tests histogram (among the people who answered the question)ggplot(aes(x = number_of_practice_tests), data= step2_complete) +geom_histogram() +theme_minimal() +xlab("Number of tests") +ylab("Frequency") +labs(title ="How many total practice tests did you take before Step 2?")
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
ggplot(aes(x = practicetest_step1), data= step1_complete) +geom_histogram() +theme_minimal() +xlab("Number of tests") +ylab("Frequency") +labs(title ="How many total practice tests did you take before Step 1?")
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
NBME and Dr. Legallo's videos and notes from preclinical
1
nbme practice exams
1
NBME practice tests
1
Online MedEd
1
Pixorize
1
Youtube
1
Code
#If you used other resources not listed above, would you use them again?step1_complete$other_hascontents <-as.integer(nchar(step1_complete$other_resources_step1) >0) #flagging the rows that have contents https://stackoverflow.com/questions/64744988/testing-to-see-if-characters-are-present-in-a-cell-in-ras_tibble(step1_complete[which(step1_complete$other_hascontents ==1), c("other_resources_step1", "other_step1") ])
# A tibble: 8 × 2
other_resources_step1 other_step1
<chr> <chr>
1 Pixorize Yes, helped immensel…
2 Dirty Medicine YouTube videos yes
3 Dirty Medicine Yes
4 nbme practice exams yes
5 NBME practice tests Yes
6 Youtube Yes
7 Online MedEd Maybe
8 NBME and Dr. Legallo's videos and notes from preclinical yes, NBME and Dr. Le…
Code
# what would you change?step1_complete$change_hascontents <-as.integer(nchar(step1_complete$changes_step1) >0) as_tibble(step1_complete[which(step1_complete$change_hascontents ==1), "changes_step1" ])
# A tibble: 39 × 1
value
<chr>
1 "I would ditch pathoma and just do Anki, UWorld, sketchy and pixorize just f…
2 "I did it in 3 weeks and it was VERY stressful. However, this ended up being…
3 "Would have started earlier in preclinical - would have ignored my preclinic…
4 "I would have studied for 4 weeks instead of 6 weeks. However, I took my tes…
5 "More SketchyMicro. Start using outside resources during 1st year. More UWor…
6 "I could have studied less (like 3-4 weeks) because I was consistently scori…
7 "I would make all of my study time part time (like I did when at home for th…
8 "I took about 3 weeks to review Sketchy Micro/Pharm and Pathoma as well as d…
9 "Focus more on the things that overlap with step 2"
10 "I just used sketchy/anki for micro, and that was great. I don't love it for…
# ℹ 29 more rows
Code
#dat$study_amount_step1_1
3 Appendix/notes
All the analyses are performed using the following:
R version 4.2.2 (2022-06-24); R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
---title: "USMLE Data Analysis"author: "Lisa Levoir"date: "`r format(Sys.time(), '%B %d, %Y')`"format: html: theme: flatly code-fold: true code-tools: true html-math-method: katex toc: true toc-depth: 3 fig-width: 13 fig-height: 10 toc-title: "Contents" number-sections: true self-contained: true self-contained-math: true smooth-scroll: true fontsize: 0.8em title-block-banner: true citation-location: margin include-after-body: graph_fold.htmleditor: visualengine: knitr---# Data import and cleaningIn this section I import the data and prepare it for analysis.```{r setup}#| warning: false#| message: false#| echo: fencedlibrary(tidyverse)library(plotly)library(dplyr)library(tidyr)library(knitr)library(table1) #Create HTML Tables of Descriptive Statistics https://cran.r-project.org/web/packages/table1/vignettes/table1-examples.html#library(OMTM1) #https://github.com/schildjs/OMTM1/library(Hmisc)library(rms) # Regression Modeling Strategies by Frank https://cran.r-project.org/web/packages/rms/index.htmllibrary(modelsummary) #Summary Tables and Plots for Statistical Models and Data: Beautiful, Customizable, and Publication-Ready https://cran.r-project.org/web/packages/modelsummary/index.htmllibrary(scales) # The scales packages provides the internal scaling infrastructure used by ggplot2, and gives you tools to override the default breaks, labels, transformations and palettes. https://scales.r-lib.orglibrary(viridis) #colorslibrary(cowplot) #allows me to use plotgridlibrary(gridExtra) #adding tables to plotslibrary(visdat) #shows missing datalibrary(GGally) #makes pairs plotslibrary(sandwich) #for robust standard errorslibrary(gtsummary) #makes the tables look nicersetwd("/Users/lisalevoir/BIOS7351_Collab/data_project2") #this line I would need to run in the consoleknitr::opts_knit$set(root.dir ="/Users/lisalevoir/BIOS7351_Collab/github/BIOS_Collaboration/project_2_analysis") #now I set global options for knitting, I also had to toggle global options > R Markdown > evaluate chunks in current directory#import the datadat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/combined_data_203.csv")#to compare that the merging went as expectedVUMSdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_VUMS.csv")HMSdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_HMS.csv")UVAdat <-read.csv("/Users/lisalevoir/BIOS7351_Collab/data_project2/DATA_UVA.csv")```### Inclusion/ExclusionCriteria to exclude students who most likely took a scored exam:- Any PhD students (n = 2)- Any 5th year program students (n = 19)- M4 students at Vanderbilt (n = 5)- Students who did not complete either Step survey (n = 2)- Students who specifically stated they took a scored Step 1 (n=1)Based on our criteria we would exclude record IDs:- VUSM: 23, 26, 39, 40, 54, 3, 8, 12, 49, 60, 62, 64- HMS: 1, 21, 28, 30, 34, 37, 39, 41, 44, 47, 49, 61- UVA: 33, 47, 80, 81, 83```{r data_cleaning}#| warning: false#| message: false#| echo: false#creating a unique ID for each participant so I can easily summarize how many people have been included/excluded at each stepdat <- dat %>%mutate(uniqueID =paste(school, record_id))VUMSdat <- VUMSdat %>%mutate(uniqueID =paste("VUSM", record_id))HMSdat <- HMSdat %>%mutate(uniqueID =paste("HMS", record_id))UVAdat <- UVAdat %>%mutate(uniqueID =paste("UVASOM", record_id))## compare the individuals that were excluded by the collaborator (in dat) vs. the full datasetall_survey_ids <-c(VUMSdat$uniqueID, HMSdat$uniqueID, UVAdat$uniqueID)collaborator_survey_ids <- dat %>%select(uniqueID) %>%pull()``````{r}#| warning: false#| message: false#| echo: false#the list of IDs we decided as a group to excludeexcludeVU <-c(23, 26, 39, 40, 54, 3, 8, 12, 49, 60, 62, 64)excludeHMS <-c(1, 21, 28, 30, 34, 37, 39, 41, 44, 47, 49, 61)excludeUVA <-c(33, 47, 80, 81, 83)'%!in%'<-function(x,y)!('%in%'(x,y)) #make a way to use the not in commandVU_in <-filter(VUMSdat, record_id %!in% excludeVU)H_in <-filter(HMSdat, record_id %!in% excludeHMS)UVA_in <-filter(UVAdat, record_id %!in% excludeUVA)#now I want to select the columns I'd like to include for all of my analysis (so they're in the proper order for a cbind). this will be relatively easy to come back to edit later, if needed. # I am also including a school identifier as a common column before I rbind all 3 school data frames togetherVU_in[,"schoolid"] <-"VU"UVA_in[,"schoolid"] <-"UVa"H_in[,"schoolid"] <-"HMS"############ plan for how I will get the data in a format I want:# - pull relevant columns by "starts with"# - confirm all column names match, then# - rbind together once# - then I can select from this sheet the questions relevant to Step 1 first with ends with "_1", and those who took Step 1 second with a "_2"############ Now pulling the common columns we're interested in as predictors and outcomes##note, for VU I removed ""number_other_courses_step1_1" and starts_with("other_courses_step1_1___1) because these questions were not on the other school surveystook_step1_VU <- VU_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step1"),starts_with("amboss_percent_step1"),starts_with("length_step1"),starts_with("practicetest_step1"),starts_with("full_test_practice_step1"), # split into binary: Yes and I am glad I did, Yes and it was unnecessary, No and I wish I did, No and I am glad I did notstarts_with("push_step1"),starts_with("push_practice_test_step1"),starts_with("push_nbme_practice_score_step1"),starts_with("push_uw_practice_score_step1"),starts_with("final_nbme_practice_score_step1"),starts_with("final_uw_practice_score_step1"),starts_with("score_step1"),starts_with("resources_step1"),starts_with("other_resources_step1"),starts_with("other_step1"),starts_with("changes_step1"),"uniqueID","schoolid", "exam_order" ) took_step1_UVA <- UVA_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step1"),starts_with("amboss_percent_step1"),starts_with("length_step1"),starts_with("practicetest_step1"),starts_with("full_test_practice_step1"),starts_with("push_step1"),starts_with("push_practice_test_step1"),starts_with("push_nbme_practice_score_step1"),starts_with("push_uw_practice_score_step1"),starts_with("final_nbme_practice_score_step1"),starts_with("final_uw_practice_score_step1"),starts_with("score_step1"),starts_with("resources_step1"),starts_with("other_resources_step1"),starts_with("other_step1"),starts_with("changes_step1"),"uniqueID","schoolid", "exam_order" )took_step1_H <- H_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step1"),starts_with("amboss_percent_step1"),starts_with("length_step1"),starts_with("practicetest_step1"),starts_with("full_test_practice_step1"),starts_with("push_step1"),starts_with("push_practice_test_step1"),starts_with("push_nbme_practice_score_step1"),starts_with("push_uw_practice_score_step1"),starts_with("final_nbme_practice_score_step1"),starts_with("final_uw_practice_score_step1"),starts_with("score_step1"),starts_with("resources_step1"),starts_with("other_resources_step1"),starts_with("other_step1"),starts_with("changes_step1"),"uniqueID","schoolid", "exam_order" )## now rbinding the three schools togethertook_step1general <-rbind(took_step1_H, took_step1_UVA, took_step1_VU)## splitting the dataset so I also have reference sheets specific to step 1 first and step 1 secondtook_step1_first <- took_step1general %>%select(ends_with("_1"), "schoolid") #I actually haven't used this data frame in the analysistook_step1_second <- took_step1general %>%select(ends_with("_2"), "schoolid") #I actually haven't used this data frame in the analysis#this function will accommodate for the survey which split up the step 2 responses by whether it was taken first or second. I would like to run analysis (and gather percent missing) across all scores as they are available. I can add an indicator column later to identify people who took it first/second. The idea with this function is the output (called storage_df) should be easy to colbind onto the original data frame. Voila!#also if we decide to change our minds and include more variables it will be quick to run.merge_my_columns <-function(input_cols, source_df){storage_df <-as.data.frame(matrix(nrow =nrow(source_df), ncol =length(input_cols)))names(storage_df) <- input_colsfor(i in1:length(input_cols)) { cols <- source_df %>%select(starts_with(input_cols[i])) %>%names()print(cols) new_cobined_col <-coalesce(source_df[, cols[1]], source_df[, cols[2]]) storage_df[, i] <- new_cobined_col}print("Above is a list of columns I have combined for you. Hope it looks right!")return(storage_df)}######## combine results across exam order for Step 2class(took_step1general$practicetest_step1_1) <-"integer"#had to change class in order to coalesce these. the original question was "How many total practice tests did you take before Step 1?" and the response field was text but it should just be integers. ## This will create a warning in the results that NAs are introduced by coercion but that is alrightclass(took_step1general$practicetest_step1_2) <-"integer"class(took_step1general$push_uw_practice_score_step1_2) <-"integer"class(took_step1general$push_uw_practice_score_step1_1) <-"integer"cols_step1 <-c("uworld_percent_step1", "amboss_percent_step1", "length_step1", "practicetest_step1","full_test_practice_step1", "push_step1", "push_practice_test_step1", "push_nbme_practice_score_step1", "push_uw_practice_score_step1" ,"final_nbme_practice_score_step1", "final_uw_practice_score_step1", "score_step1", "other_resources_step1", "other_step1","changes_step1")#now I need to handle the resources question, which was handled differentlytookstep1_resources <- took_step1general %>%mutate(I.uworld =coalesce(resources_step1_1___1 + resources_step1_2___1), I.firstaid =coalesce(resources_step1_1___2 + resources_step1_2___2),I.anki =coalesce(resources_step1_1___3 + resources_step1_2___3), I.sketchy =coalesce(resources_step1_1___4 + resources_step1_2___4), I.amboss=coalesce(resources_step1_1___5 + resources_step1_2___5), I.pathoma =coalesce(resources_step1_1___6 + resources_step1_2___6),I.boardsbeyond =coalesce(resources_step1_1___7 + resources_step1_2___7),I.other =coalesce(resources_step1_1___8 + resources_step1_2___8)) %>%select(I.uworld, I.firstaid, I.anki, I.sketchy, I.amboss, I.pathoma, I.boardsbeyond, I.other)to_add <-merge_my_columns(input_cols = cols_step1, source_df = took_step1general)#table(to_add$score_step1) ## all step 1 scores have an outcome so there is no one to drop/filter outstep1_complete <-bind_cols(took_step1general, to_add, tookstep1_resources) %>%select(record_id, uniqueID:I.other)``````{r data_cleaning_forstep2}#| warning: false#| message: false#| echo: falsetook_step2_VU <- VU_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step2"),starts_with("amboss_percent_step2"),starts_with("length_step2"),starts_with("practicetest_step2"),starts_with("full_test_practice_step2"),starts_with("practice_score_step2"),starts_with("practice_test_step2"),starts_with("score_step2"),starts_with("target_score_step2"),starts_with("resources_step2"),"uniqueID","schoolid", "exam_order" ) #note, I removed ""number_other_courses_step1_1" and starts_with("other_courses_step1_1___1) because these questions were not on the other school surveystook_step2_UVA <- UVA_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step2"),starts_with("amboss_percent_step2"),starts_with("length_step2"),starts_with("practicetest_step2"),starts_with("full_test_practice_step2"),starts_with("practice_score_step2"),starts_with("practice_test_step2"),starts_with("score_step2"),starts_with("target_score_step2"),starts_with("resources_step2"),"uniqueID","schoolid", "exam_order" )took_step2_H <- H_in %>%select(starts_with("record_id"),starts_with("uworld_percent_step2"),starts_with("amboss_percent_step2"),starts_with("length_step2"),starts_with("practicetest_step2"),starts_with("full_test_practice_step2"),starts_with("practice_score_step2"),starts_with("practice_test_step2"),starts_with("score_step2"),starts_with("target_score_step2"),starts_with("resources_step2"),"uniqueID","schoolid", "exam_order" )took_step2general <-rbind(took_step2_H, took_step2_UVA, took_step2_VU)took_step2_first <- took_step2general %>%select(ends_with("_1"), "schoolid", "uniqueID") #don't plan on using these, but available if neededtook_step2_second <- took_step2general %>%select(ends_with("_2"), "schoolid", "uniqueID")#now I need to handle the resources question, which was handled differentlytookstep2_resources <- took_step2general %>%mutate(I.uworld =coalesce(resources_step2_1___1 + resources_step2_2___1), I.firstaid =coalesce(resources_step2_1___2 + resources_step2_2___2),I.anki =coalesce(resources_step2_1___3 + resources_step2_2___3), I.sketchy =coalesce(resources_step2_1___4 + resources_step2_2___4), I.amboss=coalesce(resources_step2_1___5 + resources_step2_2___5), I.pathoma =coalesce(resources_step2_1___6 + resources_step2_2___6),I.boardsbeyond =coalesce(resources_step2_1___7 + resources_step2_2___7),I.other =coalesce(resources_step2_1___8 + resources_step2_2___8)) %>%select(I.uworld, I.firstaid, I.anki, I.sketchy, I.amboss, I.pathoma, I.boardsbeyond, I.other)########combine results across exam order for Step 2class(took_step2general$practice_score_step2_2) <-"integer"#had to change class in order to coalesce theseclass(took_step2general$score_step2_1) <-"integer"class(took_step2general$score_step2_2) <-"integer"columns_to_summarize <-c("uworld_percent_step2", "amboss_percent_step2", "length_step2", "practicetest_step2","full_test_practice_step2", "practice_score_step2", "practice_test_step2", "score_step2","target_score_step2")to_add <-merge_my_columns(input_cols = columns_to_summarize, source_df = took_step2general)#listing the people with no outcome reporteddrop_these_with_no_outcome <-bind_cols(took_step2general, to_add) %>%filter(is.na(score_step2)) %>%select("record_id", "schoolid", "uniqueID")#dropping the people with no outcomeintermediate_step2 <-bind_cols(took_step2general, to_add, tookstep2_resources) %>%filter(!is.na(score_step2))#creating the complete data setstep2_complete <- intermediate_step2 %>%select(record_id, schoolid:I.other)```There were `r length(intersect(all_survey_ids, collaborator_survey_ids))` participants after the the clients did data exclusion by hand, as compared to the total survey size which was `r length(step1_complete$uniqueID)` unique individuals included in the step 1 data.These are the record IDs that were excluded by the collaborator: `r setdiff(all_survey_ids, collaborator_survey_ids)`.For our analysis, I originally included everyone from the 3 school specific data sets. Then, we flagged individuals to remove and I remade the source data set. These are the records IDs that we decided to exclude:- VU record ids: 23, 26, 39, 40, 54, 3, 8, 12, 49, 60, 62, 64- HMS record ids: 1, 21, 28, 30, 34, 37, 39, 41, 44, 47, 49, 61- UVA record ids: 33, 47, 80, 81, 83Below are tables for Step 1 and Step 2 response inclusion by school:```{r}kable(table(step1_complete$schoolid), caption ="Number of Step 1 responses included by school")kable(table(step2_complete$schoolid), caption ="Number of Step 2 responses included by school")```Notice that unfortunately for Step 2 analysis, we had to drop `r nrow(drop_these_with_no_outcome)` individuals who did not report a step 2 score. These raw counts and frequencies of people who did not give a step 2 score (and are therefor not eligible for analysis) are listed by institution below.```{r}# describing the missingnesskable(table(drop_these_with_no_outcome$schoolid), caption ="Number of missing Step 2 scores by institution")num <-as.vector(table(drop_these_with_no_outcome$schoolid))denom <-as.vector(table(step2_complete$schoolid))nonresponse_freq <-setNames(c(round(num/denom, 3)), c(names(table(step2_complete$schoolid))))kable(nonresponse_freq, caption ="frequency of missing step 2 scores by instition")```We also find it helpful to visually profile the missingness in this graphic below. Some of this missingness is due to questions being hierarchical.```{r}######### visually profile missing responsesp1 <- visdat::vis_miss(step1_complete)p2 <- visdat::vis_miss(step2_complete)title_gg <-ggdraw() +draw_label("Response missingness for pooled survey results across exam order")gridded <-plot_grid(p1, p2, label_size =12, ncol =2, align ="hv", label_x =0.5,labels =c("Step 1","Step 2"))plot_grid(title_gg, gridded, nrow =2, rel_heights =c(0.1, 0.9))```# AnalysisNote, we plan to use the robust/sandwich variance estimator for regression models. One inclusion criteria is that the outcome variable ("Y") must be available for a subject to be included in the analysis question (ie. if they did not report a step 2 score, we won't perform relevant step 2 analysis on them).::: panel-tabset## Q1 Does order have an impact on Step 2 scores?We cannot analyze Step 1 since all survey responses reported passing For Step 2 scores, I will perform a linear regression with:- Y = Step 2 score- X = factor ( 1 = "step 1 first", 2 = "step 2 first", 3 = "only step 1", 4 = "only step 2")- Z = school (need to adjust for this)```{r exam_order_analysis}step2_complete[ ,"order_factor"] <-factor(step2_complete$exam_order, levels =c(1,2, 3, 4), labels =c("step 1 first", "step 2 first", "only step 1", "only step 2"))step2_complete[ ,"which_exam_first"] <-ifelse(step2_complete$order_factor =="step 1 first"| step2_complete$order_factor =="only step 1", 1, 2)step2_complete[ ,"school_factor"] <-factor(step2_complete$schoolid, labels =c("HMS", "UVA", "VUMS"))table(step2_complete$order_factor)table(step2_complete$which_exam_first)exam_order_mod <-lm(formula = score_step2 ~ which_exam_first + school_factor, data = step2_complete)sandwich(exam_order_mod) #this gives the sandwich variancesummary(exam_order_mod) #model summary for reporting```Based on the model output (p values), there doesn't appear to be any signifigant associations between exam order and the score for Step 2. Since the $R^2$ value is essentially 0, I conclude that there was no effect of exam order on step 2 scores.## Q2 What factors affect Step 2 score?Again, we cannot analyze Step 1 scores since all respondents reported passing.Based on our SAP, if there are any covariates with more than 30% of responses missing, we will drop that variable or populate it with 0, depending on context. For example, the percent of Amboss questions completed will be filled with 0 for people who didn't answer that they used Amboss since it seems safe to assume they didn't complete any of the Amboss questions. For people who did select that they used Amboss but didn't answer a percentage will be treated as missingIf less than 30% are missing, I may consider performing bootstrap sampling of known values to replace missing values.After accounting for missingness, I will assess for co-linearity of the predictors (ie. correlation) using VIF. If there is high co-linearity, we will use LASSO to perform variable selection. If there is no evidence of concerning levels of colinearity, I will proceed with linear regression.```{r data_profile_and_pairs_plots}#| warning: false#here I am creating the cleaned Uworld and Amboss variables. if they did not check that they used it, the percentage is set to 0.# if they did use it then the next ifelse checks if there is an NA where there should be a percent# if there is an NA where there should be a percent, we maintain this NA. If there is not, then we can report the percentage.# same process for uworldstep2_complete <- step2_complete %>%mutate(amboss_clean =ifelse(I.amboss ==0, 0, ifelse(is.na(step2_complete$amboss_percent_step2) ==TRUE, NA, step2_complete$amboss_percent_step2)),uworld_clean =ifelse(I.uworld ==0, 0, ifelse(is.na(step2_complete$uworld_percent_step2 ) ==TRUE, NA, step2_complete$uworld_percent_step2)))######## profile missingness in the step 2 data and addresspercents_missing <-round(colSums(is.na(step2_complete))/nrow(step2_complete), 3)*100kable(percents_missing, caption ="Percent missing observations for pooled Step 2 survey")#inspecting percent missing, it seems like most responses are now complete enough for analysisstep2_complete$amboss_percent_step2 <-ifelse(is.na(step2_complete$amboss_percent_step2) ==TRUE, 0, step2_complete$amboss_percent_step2)class(step2_complete$practicetest_step2) <-"integer"step2_complete[,"on_target"] <-factor(step2_complete$target_score_step2, levels =c(1,2,3), labels =c("at target", "above target", "below target"))step2_complete[, "practice_test_2_clean"] <-ifelse(is.na(step2_complete$practicetest_step2) ==TRUE, NA, substr(step2_complete$practicetest_step2, start =1, stop =1))step2_complete[, "number_of_practice_tests"] <-as.numeric(step2_complete$practice_test_2_clean)```Multiple linear regression with:- Y = Step 2 score- X1 = % UWorld- X2 = % Amboss- X3 = length study- X4 = \# of practice tests- X5 = full test day (yes/no code as binary)- X6 = final practice score (however, is there some conversion between U World and Amboss - Jeffrey is looking into this)- Z = School (need to adjust for this)Careful to note that not all cases are complete - for example there are 399 responses in the complete step 2 dataset, of which for the number of practice tests taken, 108 are missing and 291 have a response recorded.Below I report the model results, sandqich variance, and VIF for step 2 scores model.```{r}step2_complete$simulate_full_practice <-ifelse(step2_complete$full_test_practice_step2 <=2, 0, 1)step2_complete$simulate_full_practice <-factor(step2_complete$simulate_full_practice, levels =c(0, 1), labels =c("Yes", "No"))step2_complete$length_step2 <-factor(step2_complete$length_step2 , levels =c(1:6), labels =c("less than 1 week", "1-2 weeks", "3-4 weeks", "5-6 weeks", "7-8 weeks", "more than 8 weeks"))mod_step2_scores <-lm(score_step2 ~ uworld_percent_step2 + amboss_percent_step2 + length_step2 + simulate_full_practice + practice_score_step2 + number_of_practice_tests + school_factor, data = step2_complete)summary(mod_step2_scores)sandwich(mod_step2_scores)vif(mod_step2_scores)mod_step2_scores2 <-lm(score_step2 ~ uworld_percent_step2 + amboss_percent_step2 + length_step2 + simulate_full_practice + practice_score_step2 + practice_test_2_clean + school_factor, data = step2_complete)summary(mod_step2_scores2) #we need to talk about these model results```I looked into the difference between practice_test_step2 (the final practice test I took before my exam was... 8 options) and practicetest_step2 (text response of how many practice tests did you take before step 2). I decided not to include either in the model.::: callout-noteWe need to talk about the model results between the two ways I coded the variables for \# of practice tests:::## Q3 What is associated (in this data) with pushing back a Step 1 exam date?Here, I will perform logistic regression withY = yes or no (1 = yes, 2 = no for "push_step1")There may not be sufficient data on this since only 20 people responded that they decided to push back Step 1. The factors that were measured are:- push remember step1 (1 = I only remember the form name, 2 = I only remember the score, 3 = I remember the form name and the score, 4 = I don't remember either) - we decided not to include this variable (can change later if desired)- push score only step 1 (1 = NBME, 2 = Uworld)- push practice test step 1 ( 1 - 8 listing various exams)- push nbme practice score (from 0 to 100%)- push uw practice score (from 180 to 300)Listing variables by name and if I have included them:- "push_step1_1" yes- "push_remember_step1_1" not included b/c a precursor question- "push_score_only_step1_1" not currently included but could be- "push_practice_test_step1_1" yes- "push_nbme_practice_score_step1_1" yes- "push_uw_practice_score_step1_1" yes```{r}step1_complete$push_step1 <-ifelse(step1_complete$push_step1 ==2|is.na(step1_complete$push_step1) ==TRUE, 2, 1) #recording the NA's to be "No" (they did not push back step 1)step1_complete$push_step1_label <-factor(step1_complete$push_step1, levels =c(1, 2), labels =c("Yes", "No")) #making a nice descriptive labeldid_push_df <- step1_complete %>%filter(push_step1_label =="Yes")dat %>%filter(!is.na(push_remember_step1_1))step1_complete$push_practice_test_step1 <-factor(step1_complete$push_practice_test_step1, levels =c(1:8), labels =c("NBME 25", "NBME 26", "NBME 27", "NBME 28", "NBME 29", "NBME 30", "UWorld 1", "UWorld 2"))units(step1_complete$push_uw_practice_score_step1) <-"score units"units(step1_complete$push_nbme_practice_score_step1) <-"percent"label(step1_complete$push_practice_test_step1) <-"exam that triggered pushing back"label(step1_complete$push_nbme_practice_score_step1) <-"NBME P(passing) that triggered pushing back"label(step1_complete$push_uw_practice_score_step1) <-"3 digit UWorld score that triggered pushing back"caption <-"Description of indiviuals that pushed back Step 1"table1(~ push_practice_test_step1 + push_uw_practice_score_step1 + push_nbme_practice_score_step1 |push_step1_label, data=step1_complete, topclass="Rtable1-zebra")```## Descriptive Statistics (still in progress)Descriptive statistics will be reported by school and total.- histogram of Step 2 scores- what resources are most widely used barplot- how long did students study barplot- number of practice tests histogram (among the people who answered the question)- summarize comments (other_resources, other_step, changes_step)```{r}## do a plot for all schools (in total) as well as Vanderbilt in particular## use a unified color schemeggplot(aes(x = score_step2), data= step2_complete) +geom_histogram() +theme_minimal() +xlab("Self Reported Step 2 Score") +ylab("Frequency") +labs(title ="Frequency of reported Step 2 Scores")#more data cleaning first first step 2 for the resouces they selected that they usedresources_step2 <-colSums(took_step2general[20:35])uworld <- resources_step2[1] + resources_step2[9]first_aid <- resources_step2[2] + resources_step2[10]anki <- resources_step2[3] + resources_step2[11]sketchy <- resources_step2[4] +resources_step2[12]amboss <- resources_step2[5] + resources_step2[13]pathoma <- resources_step2[6] + resources_step2[14]boards_and_beyond <- resources_step2[7] + resources_step2[15]other <- resources_step2[8] + resources_step2[16]totals <-c(uworld, first_aid, anki, sketchy, amboss, pathoma, boards_and_beyond, other)#same idea for step 1resources_step1 <-colSums(took_step1general[26:41])uworld1 <- resources_step1[1] + resources_step1[9]first_aid1 <- resources_step1[2] + resources_step1[10]anki1 <- resources_step1[3] + resources_step1[11]sketchy1 <- resources_step1[4] +resources_step1[12]amboss1 <- resources_step1[5] + resources_step1[13]pathoma1 <- resources_step1[6] + resources_step1[14]boards_and_beyond1 <- resources_step1[7] + resources_step1[15]other1 <- resources_step1[8] + resources_step1[16]totals1 <-c(uworld1, first_aid1, anki1, sketchy1, amboss1, pathoma1, boards_and_beyond1, other1)rdf <-data.frame(amount =c(totals, totals1), name =rep(names(totals), 2), step_exam =c(rep("Step 2", 8), rep("Step 1", 8) ))# http://www.sthda.com/english/wiki/ggplot2-barplots-quick-start-guide-r-software-and-data-visualization used this as a guide#what resources are most widely used? I want to sort this barplot in order of frequency but for some reason this wasn't working for me with reorder()ggplot(data=rdf, aes(x=name, y=amount, fill = step_exam)) +geom_bar(stat="identity", position=position_dodge())+theme_minimal() +coord_flip() +scale_fill_brewer(palette="Blues")## how long did the students study barplot# for step 1step1_complete$length_step1 <-factor(step1_complete$length_step1 , levels =c(1:6), labels =c("less than 1 week", "1-2 weeks", "3-4 weeks", "5-6 weeks", "7-8 weeks", "more than 8 weeks"))ggplot(data.frame(step1_complete), aes(x=length_step1)) +geom_bar() +theme_minimal() +xlab("Time") +ylab("Frequency") +labs(title ="How long did you study for Step 1 during a protected study period?")# for step 2ggplot(data.frame(step2_complete), aes(x=length_step2)) +geom_bar() +theme_minimal() +xlab("Time") +ylab("Frequency") +labs(title ="How long did you study for Step 2 during a protected study period?")## number of practice tests histogram (among the people who answered the question)ggplot(aes(x = number_of_practice_tests), data= step2_complete) +geom_histogram() +theme_minimal() +xlab("Number of tests") +ylab("Frequency") +labs(title ="How many total practice tests did you take before Step 2?")ggplot(aes(x = practicetest_step1), data= step1_complete) +geom_histogram() +theme_minimal() +xlab("Number of tests") +ylab("Frequency") +labs(title ="How many total practice tests did you take before Step 1?")## summarize comments (other_resources, other_step, changes_step)#"other_resources_step1" "other_step1" "changes_step1" kable(table(step1_complete$other_resources_step1), caption ="Other listed resources for Step 1")#If you used other resources not listed above, would you use them again?step1_complete$other_hascontents <-as.integer(nchar(step1_complete$other_resources_step1) >0) #flagging the rows that have contents https://stackoverflow.com/questions/64744988/testing-to-see-if-characters-are-present-in-a-cell-in-ras_tibble(step1_complete[which(step1_complete$other_hascontents ==1), c("other_resources_step1", "other_step1") ])# what would you change?step1_complete$change_hascontents <-as.integer(nchar(step1_complete$changes_step1) >0) as_tibble(step1_complete[which(step1_complete$change_hascontents ==1), "changes_step1" ])#dat$study_amount_step1_1```:::# Appendix/notesAll the analyses are performed using the following:- R version 4.2.2 (2022-06-24); R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.- Harrell Jr FE (2022). rms: Regression Modeling Strategies. R package version 6.3-0, <https://CRAN.R-project.org/package=rms>.The table below lists packages used in this document.```{r}subset(data.frame(sessioninfo::package_info()), attached==TRUE, c(package, loadedversion))```
